Re: [OE-core][PATCH v6 09/15] spdx30: Skip install package CVE information

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Thu, 2026-03-12 at 08:15 -0600, Joshua Watt wrote:
> On Thu, Mar 12, 2026 at 5:55 AM Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> > 
> > On Tue, 2026-03-10 at 12:38 -0600, Joshua Watt via lists.openembedded.org wrote:
> > > Skips adding the install package CVE information by default. This
> > > information grows exponentially, since it ends up be N_CVES *
> > > N_PACKAGES. The CVE information for a given installed package can be
> > > determined by following the "generates" link between the install package
> > > and the recipe and looking at the CVE information for the recipe,
> > > meaning that the CVE information is only included once in the SPDX
> > > document.
> > > 
> > > If users still need the legacy method of including CVE information for
> > > each package, then then can set SPDX_PACKAGE_INCLUDE_VEX = "1"
> > > 
> > > Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
> > > ---
> > >  meta/classes/create-spdx-3.0.bbclass | 11 ++++++++
> > >  meta/lib/oe/spdx30_tasks.py          | 39 ++++++++++++++--------------
> > >  meta/lib/oeqa/selftest/cases/spdx.py | 12 +++++++++
> > >  3 files changed, 43 insertions(+), 19 deletions(-)
> > > 
> > > diff --git a/meta/classes/create-spdx-3.0.bbclass b/meta/classes/create-spdx-3.0.bbclass
> > > index c3ea95b8bc..88b7ef9f42 100644
> > > --- a/meta/classes/create-spdx-3.0.bbclass
> > > +++ b/meta/classes/create-spdx-3.0.bbclass
> > > @@ -45,6 +45,17 @@ SPDX_INCLUDE_VEX[doc] = "Controls what VEX information is in the output. Set to
> > >      including those already fixed upstream (warning: This can be large and \
> > >      slow)."
> > > 
> > > +SPDX_PACKAGE_INCLUDE_VEX ?= "0"
> > > +SPDX_PACKAGE_INCLUDE_VEX[doc] = "Link VEX information to the binary package outputs. \
> > > +    Normally, VEX information is only linked to the common recipe that `generates` the \
> > > +    binary packages, but setting this to '1' will cause it to also be linked into the \
> > > +    generated binary packages. This is off by default because linking the VEX data to \
> > > +    each package causes the SPDX output to grow very large, and the same information \
> > > +    can be determined by following the `generates` relationship back to the recipe. \
> > > +    Before recipe packages were introduced, this was the only way VEX data was \
> > > +    expressed; you may need to enable this if your downstream tools do not \
> > > +    understand how to trace back to the recipe to find VEX information."
> > 
> > To me, removing this duplication and keeping the SPDX docs usable seems
> > like a very sensible thing to do. Do we want/need to make it
> > configurable?
> > 
> > I appreciate some tools/usage may need fixing to work with this but
> > adding configuration options like this makes it harder to use our code
> > and also adds maintenance/testing overhead.
> 
> Maybe, but I'm very hesitant to break any existing SPDX based CVE
> workflows that people may have in a LTS release, which is the only
> reason I added the option. I'm fine to remove this after the LTS, IMHO
> it's just too close to LTS release to suddenly say "sorry this is all
> broken for you now"

We're not yet at feature freeze and I'm getting very worried about all
the combinations of things we're trying to support. Having so many user
options is likely going to confuse users too . I can see both sides of
this but I am leaning towards a cleaner implementation and having
people adapt to it now rather than when we remove it later.

Do we know of tools that are going to struggle with this change? I'm
guessing people can adapt to the change relatively easily?

Cheers,

Richard